Comparison of methods to model aboveground biomass and derive 20+ year aboveground biomass trajectories
Scott
L
Powell, USDA Forest Service, Pacific Northwest Research Station, scott.powell@oregonstate.edu
(Presenting)
Robert
E
Kennedy, USDA Forest Service, Pacific Northwest Research Station, robert.kennedy@oregonstate.edu
Sean
P
Healey, USDA Forest Service, Rocky Mountain Research Station, seanhealey@fs.fed.us
Ken
B
Pierce, WA Dept. of Fish and Wildlife, pierckbp@dfw.wa.gov
Warren
B
Cohen, USDA Forest Service, Pacific Northwest Research Station, warren.cohen@orst.edu
Gretchen
G
Moisen, USDA Forest Service, Rocky Mountain Research Station, gmoisen@fs.fed.us
Janet
L
Ohmann, USDA Forest Service, Pacific Northwest Research Station, johmann@fs.fed.us
Spatially explicit knowledge of aboveground biomass dynamics at broad scales is critical to understanding how forest disturbance and regrowth processes influence carbon cycling. In support of the North American Carbon Program (NACP) and the Forest Inventory and Analysis (FIA) program, we linked FIA field data with 20+ year time-series of radiometrically normalized Landsat satellite imagery to derive trajectories of aboveground forest biomass for two distinct study locations in Arizona and Minnesota. We compared a variety of statistical techniques for modeling biomass at a single point in time, to better understand the effects of predictor variable inclusion. In addition to spectrally-derived predictor variables, the inclusion of topographic and climatic variables yielded only incremental gains for the non-parametric model types. The “best” model from each technique was then applied to each date in the 20+ year Landsat time-series to derive preliminary biomass trajectories. Finally, a curve-fitting algorithm was applied to the preliminary biomass trajectories to leverage the temporal information contained within the time-series itself and to minimize error associated with year-to-year variability in biomass predictions. The effect of the curve-fitting algorithm was a pronounced improvement on predictions of biomass change when validated against observed biomass change from repeat field inventories. The application of these techniques to a large sample of Landsat scenes across North America will enable spatial and temporal estimation of biomass dynamics associated with forest disturbance and regrowth.
NASA Carbon Cycle & Ecosystems Active Awards Represented by this Poster: